Modeling and implementation of solder-activated joints for
single-actuator, centimeter-scale robotic mechanisms
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Telleria, Maria J, Malik Hansen, Don Campbell, Amelia Servi,
and Martin L Culpepper. Modeling and Implementation of Solderactivated Joints for Single-actuator, Centimeter-scale Robotic
Mechanisms. In Pp. 1681–1686, 2010. © Copyright 2010 IEEE
As Published
http://dx.doi.org/10.1109/ROBOT.2010.5509720
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Version
Final published version
Accessed
Thu May 26 09:04:33 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/78616
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
2010 IEEE International Conference on Robotics and Automation
Anchorage Convention District
May 3-8, 2010, Anchorage, Alaska, USA
Modeling and Implementation of Solder-activated Joints for SingleActuator, Centimeter-scale Robotic Mechanisms
Maria J. Telleria1, Malik Hansen2, Don Campbell2, Amelia Servi1, and Martin L. Culpepper1

Abstract— We explain when, and why, solder-based phase
change materials (PCMs) are best-suited as a means to modify
a robotic mechanism’s kinematic and elastomechanic behavior.
The preceding refers to mechanisms that possess joints which
may be thermally locked and unlocked via a material phase
change within the joint. Different combinations of locked and
unlocked joints can yield several one-DOF mechanisms states.
One actuator may be used to control motion allowed by a first
state, then a new combination of locked/unlocked joints may be
set and the actuator then controls motion allowed by the new
state. Compared to other thermo-rheological fluids, solders
yield joints with the (i) highest strength and stiffness, (ii) fastest
lock/unlock speed, and (iii) lowest lock/unlock power. Herein,
we cover physics-based design insights that provide
understanding of how solder-based material properties and
joint design dominate/limit joint performance characteristics.
First order models are used to demonstrate selection of suitable
PCMs and how to set initial joint geometry prior to fine tuning
via detailed models/experiments. The insights and models are
discussed in the context of a joint for a crawling robot that uses
a single spooler motor and three solder-locking joints to crawl
and steer.
A
I. INTRODUCTION
most troublesome problem in multi degree-of-freedom,
cm-scale robot design is the need for multiple sub-cm
actuators. It is often impossible to find off-the-shelf
actuators that exhibit the requisite force/torque, power and
speed characteristics in a sub-cm package. Even if one
could find suitable actuators, there are other practical issues
associated with miniaturization. For example, it is difficult
to create suitably small, stiff and strong mounting points for
many actuators upon a mechanism. Also, the packaging of
requisite electronics and routing of power lines to multiple
actuators is non-trivial.
These problems may be avoided when a robot’s
mechanism possesses joints that may be thermally locked
and unlocked via a phase change material (PCM) within the
joint. Different combinations of locked and unlocked joints
yield several one-DOF mechanism states. One actuator may
be used to control motion allowed by a first state, then a new
combination of locked/unlocked joints may be set and the
actuator then controls motion allowed by the new state. Our
example robot, Squishbot1 shown in Fig. 1, was designed to
be 1.5cm in diameter when extended. It contains 3 solder
locking joints (see inset) and one DC motor spooling
mechanism. This actuator-joint combination enables states
that permit locomotion along the robot’s axis, side-to-side
steering and lifting of the front (rightmost) half of the robot.
The use of phase change materials, a thermo-rheological
(TR) fluid, for mechanism fixation has been examined
within MEMS mechanisms and micro-optical alignment [1].
In these prior works, the mechanism was fixated by the
PCM, thereby enabling position/orientation holding without
powered actuation. This differs from our approach as it
does not change the mechanism’s kinematic or
elastomechanic behavior.
The scope of this paper
3
1
2
4
Fig 1. Squishbot1 – A single actuator robot. The first inset shows three solder-locking joints. The second shows a solid model of one of the rolling flexure
locking joints to wherein u-shaped copper elements (1 & 4) are attached to the sides of the top and bottom joint elements. The flexure joint, a Jacob’s
ladder joint, uses flexures to keep the top and bottom half-cylinders in contact. Rolling of the top half-cylinder over the bottom is constrained when the
solder (2) between their side plates is solidified. Current is run through a strain gage (3) within the joint to melt the solder.
Manuscript received September 15, 2009. This work was supported in
part by the U.S. Defense Advanced Research Projects Agency (DARPA)
under the Chemical Robots Program.
1
Department of Mechanical Engineering, Massachusetts Institute of
Technology, Cambridge, MA 02139. (email: mtelleri@mit.edu,
culpepper@mit.edu).
2
Boston
Dynamics,
Waltham,
MA.02451
(email:
playter@bostondynamics.com)
978-1-4244-5040-4/10/$26.00 ©2010 IEEE
encompasses the use of PCMs that lock and unlock joints,
thereby altering the mechanism’s kinematic and
elastomechanic behavior. The use of PCM joints was first
suggested and experimentally verified via bench-level
experiments (using hot glues) by Boston Dynamics.
1681
This approach enables a new means of achieving complex
motions/tasks with small-scale robotics, but it is not a
panacea.
Key performance characteristics – strength,
stiffness, speed, weight – depend upon proper PCM
selection and component design. The most salient issue is
power. The phase change process can be prohibitively
power-intensive and slow if one does not select a suitable
PCM and properly design the thermal and mechanical
behavior of the components that supply and withdraw heat
from the material.
The robot/joint design problem is multi-domain and
highly coupled, therefore modeling + experimentation rather
than intuition + iteration are the fastest means to achieve
useful performance. Herein we provide the basis to
understand (i) the physics that dominate and limit joint
performance, (ii) how to select the PCM, and (iii) how to
rapidly use 1st order models to set an initial joint geometry
for specific performance characteristics. We use the
preceding to show that solder-based PCMs are generally the
best for applications that require joints which possess the (i)
highest strength and stiffness, (ii) fastest lock/unlock speed,
and (iii) lowest lock/unlock power. The preceding is put in
context via discussion of application to Squishbot1.
A. Squishbot Performance Requirements and Constraints
Squishbot1, created during the DARPA Chemical Robots
Challenge, is a step toward the creation of small-scale, soft
robots that can deform to gain access to spaces with small
openings. Table I provides a summary of the functional
requirements and constraints encountered in the design.
TABLE I REQUIREMENTS AND CONSTRAINTS
Power supply
Minimum range
Minimum speed
Cross section size
2 - 150mAh batteries+
5 meters
4.2 mm/sec
Travel through 1cm hole
+ Li-ion polymer battery, model GMB051235 from the Guangzhou
Markyn battery company.
A key driver in the design was the need to travel through
a 1 cm hole. This requirement placed difficult size
constraints upon mechanism, actuator, on-board electronics
and power source. Power limitations required the reduction
of robot weight. These weight limits and difficulties in
finding and integrating suitable sub-cm actuators drove the
decision to use one actuator with locking/unlocking
mechanism joints.
II. BACKGROUND
In general, joint locking and unlocking may be achieved
with a variety of ‘active fluids’, i.e. materials that can
change from liquid to a solid or semi-solid state. This paper
focuses on PCMs; however we provide an overview of
competing materials as a basis for arguing why solder-based
PCMs are well-suited for small-scale robotics applications.
A. Overview of Active Fluids
Active fluids possess tunable rheological properties. They
are classified by the means used to cause a change in state
within the fluid. For example, the resistive shear stress
within magneto-rheological (MR) fluids is proportional to
the magnetic field within the fluid. Failure stresses of 50kPa
(7.25 psi) have been reported [2]. These materials have been
used in vibration dampers, and locking spherical joints [2,3].
The failure stress within electro-rheological (ER) fluids is
controlled by the electric field within the fluid. Failure
stresses of 130 kPa (18.9 psi) have been reported [4]. The
imposition of a field in MR and ER fluids forces nanoscale
particles to align within the fluid thereby enabling the fluid
to resist deformation [2,4]. This alignment mechanism
endows them with millisecond response times but places
fundamental limits on failure stresses (at practical field
strengths). Their low failure stresses limit the strength of a
locked joint and therefore the strength of the mechanism.
The change in state of photo-rheological (PR) fluids is
caused by exposure to ultra violet light. This induces a
chemical change within the fluid that leads to a change in
deformation resistance characteristics. Viscosity changes of
4 orders of magnitude have been reported, however the
highest absolute viscosity is comparable to that of honey [5].
For robotic applications, the fluid must be reversible, i.e.
switch between ‘soft’ and ‘hard’ states. Most PR fluids are
non-reversible, meaning they cannot be returned to their
original ‘soft’ state once activated. There are a limited
number of newly-available reversible PR fluids; however
their response time is slow, e.g. 4 minutes [6]. A favorable
characteristic of PR fluids is that they do not require
constant energy input to maintain a locked state, unlike MR
and ER.
Thermo-rheological (TR) fluids achieve a change in state
via temperature change. For practical purposes, we limit our
discussion to those which change phase within temperatures
that are reasonable for robotics applications, e.g. within a
few 100°C of room temperature. We consider TR fluids to
be in their ‘off-state’ at room temperature.
Common TR fluids - solders (~ 40 MPa), and hot glues (~
5 MPa) - exhibit failure stresses that are well above those of
MR, ER and PR fluids. For a given joint geometry, they
will provide higher strength; however this is ‘for naught’ if
the fundamental issues governing the (i) energy
requirements and (ii) lock/unlock speed are not addressed.
Phase change is inherently energy intensive, which is
problematic given the limited energy that may be stored onboard a small robot.
The thermal characteristics of the PCM – specific heat,
latent heat and melting temperature – must be selected so as
to minimize melting energy. This is critical as there are no
suitable small-scale heat recovery devices, e.g.
thermoelectrics, therefore melting energy is unrecoverable.
It is important to note that the limits of solidification/lock
and melting/unlock speeds (equate to robot speed) are
defined in-part by thermal characteristics. Melt time is
inversely proportional to power (which is in limited supply)
1682
and the brevity of cooling time is determined in-part by the
PCM’s thermal diffusivity (varies widely between TR
fluids) and the free convection of heat out of the PCM.
The Pugh chart in Table II contains a qualitative
comparison of the different types of active fluids. The table
contains two types of TR fluids to better represent the wide
range in properties that may be seen in TR fluids.
compression, and combinations are possible. The shear and
tension modes are shown in a generalized geometry within
Fig. 2.
Shear
Solder
Tension
Base
Material
TABLE II PUGH ACTIVE FLUID COMPARISON CHART BASED ON COMMON
CONSIDERATIONS IN ROBOTIC DESIGN+ [2,4-9].
+
Speed
Weight
Strength
Power
Scalability
PR
0
0
0
0
0
MR
++
+
0
--
ER
++
+
+
0
-
Wax
0
+
+
0
0
Solder
+
++
++
+
0
Fig. 2. Generalized shear and tension solder joint geometries.
PR is used as a baseline for qualitative comparison
B. The Case for Solder as an Active Fluid in Robotics
The strength of solder in its locked state is at least an
order of magnitude larger than that of the other active fluids.
Solder has a high thermal diffusivity, typically an order of
magnitude, or larger, than waxes and glues. This enables
solder to quickly spread thermal energy throughout its
volume thereby speeding melting and solidification. The
relevant thermal property, thermal diffusivity, is the key to
obtaining short response time in TR fluids. Glues and
waxes have low diffusivity and are therefore slow to
melt/solidify. In contrast, anyone who has soldered wires for
electronics knows how rapidly solders may be melted and
cooled.
The electronics packaging community has created many
different solders, and characterized/optimized their melt
energy, melt times and solidification times. Therein lays a
wealth of knowledge regarding material properties of
solders and guidance on customizing properties. When
competing thermal and mechanical requirements are not
satisfied by off-the-shelf solders, it is possible to modify
their composition and obtain the desired characteristics.
Squishbot1 uses solder as its locking fluid. The high
failure stress of the solder was necessary given the small
area available for locking in the mm-scale joints. The
mechanical and thermal elements of the joints were easily
scaled down to mm-size. The mechanical components are
3D printed, laser cut and/or CNC micromilled. Heat is easily
delivered to the joints via current routed through small strain
gauges that act as heaters. A perfect solder was not found
within off-the-shelf solders, therefore we selected one that
was close and therefore easily modified to minimize (i) the
amount of energy needed to unlock the joints and (ii) the
lock/unlock cycle times.
III.
JOINT DESIGN
A. Joint Mechanical Modeling
There are several concepts for joint designs where locking
could be used. We classify them by loading mechanism.
Shear and tension are the most common, but torsion,
The strength of a TR joint may be calculated using the
appropriate failure stress (shear or tension) of the solder and
the surface area of the joint. Equation (1) links failure load
to failure stress via the area over which the solder resists
load.
(1)
Fload  A  
The failure stresses of solders are readily obtained from
standard electronics handbooks [7-9] or from manufacturers.
Most solders are metal alloys, and changing the mass
fraction of alloying elements can induce large changes upon
thermal and mechanical properties. Other properties of the
joint may affect strength. Changes in the gap (ideally 10100 micrometers) may affect the strength of the joint as
detailed in [7]. It is also known that one must consider the
characteristics of the base material, for example coppertin/lead joints generally have the highest shear strength. The
joining and operating temperatures may also affect strength
[8]. The number of factors that may affect strength are
many, therefore it is typically best to use standard failure
stress data to obtain a first approximation for joint strength
and follow best practices [7-9] for other factors. If other
factors become relevant, experimentation is used to quantify
effect on performance.
B. Joint Thermal Modeling
Complex thermal models are time and resource intensive,
especially for phase change and transient problems. In
locking joints, model accuracy depends upon the accuracy
of convection coefficients, which may vary with joint design
and joint orientation during operation. As a result; FEAtype models have limitations. It is possible to use first
principles to obtain rapid, practical estimates of melting
energy, melt time, and solidification time. This information
may be used to set the initial design of a joint, which may
then be finely tuned via experiments or more detailed
thermal modeling. Our focus in this section is the means
one can use to (i) understand how performance scales with
geometry/material properties, (ii) select desired materials
and (iii) set initial design parameters for joint geometry.
A convenient way to approach modeling heat transfer is
modeling via thermal circuits where each component of the
joint has a thermal resistance. Equation (2) calculates the
conductive thermal resistance, Rcond, where k is the thermal
conductivity of the material, L is the length the heat must
1683
travel and Ak is the area perpendicular to the heat flow.
(2)
Rcond  L
Ak
k
Equation (3) is used to estimate convective thermal
resistance, Rconv, where h is the thermal convection
coefficient (~10W/m2 in air) and Ah is the surface area
exposed to the air.
(3)
Rconv  1
Ah h
The act of thermal resistance modeling forces one to think
in terms of joint geometry and material properties that
govern heat flow. A high thermal resistance indicates a low
heat transfer rate. The heat transfer rate, q, depends on
thermal resistance and temperature difference, ∆T:
(4)
q  T .
R
The first step is to ‘break’ the joint into elements of a
thermal circuit. Fig. 3 shows the thermal circuit for half of a
joint in Squishbot1. Heat flows from the heater through the
copper and solder layers via conduction before dissipation to
the ambient via convection. Heat also flows from the heater
into the Teflon via conduction.
Rconvection Rcopper-solder
q
Rteflon
Heater
Fig. 3. Half Squishbot1 joint with corresponding thermal circuit
Next we find the equivalent resistance for each of these
paths by adding the resistances as in electric circuit
modeling. As Table III shows, in Squishbot1 the low
thermal resistance (Degree Kelvin per Watt) of the copper
and solder layers indicates that these will quickly adjust
temperature. The high convection resistance (due to the
product of small surface area and small convection
coefficient) indicates low heat transfer rate to the
environment. There is less resistance to heat that travels
through the Teflon than heat that is convected to the
environment. This is important to consider, particularly for
the cooling time.
TABLE III THERMAL RESISTANCES FOR COMPONENTS OF SQUISHBOT1 JOINT
Rconvection
Rcopper-solder Rteflon
Resistance (K/W) 4.34x103
2.15x10-3
3.68x102
The lumped thermal capacity model is used to simplify
transient analysis. This is a valid assumption if the Biot
number of a component is less than 0.1, which occurs when
the temperature within the system being modeled will not
differ from that at the surface by more than 5%. The Biot
number is a ratio of the internal conduction resistance to
external convection resistance, as shown by (5) [10].
(5)
Biot  hL
k
In Table VI we can see that the Biot number is less than
0.1 for all the components of the Squishbot1 joint, the
lumped model is appropriate. Another parameter to consider
is the characteristic time for conduction within each
component. Equation (6) shows the formula used to
calculate the characteristic time, which depends upon the
distance heat must travel, L, and the material’s diffusivity, α.
2
(6)
tc  L

Table IV lists the characteristic times for each component,
which indicate how quickly a component will adjust its
temperature.
TABLE IV JOINT PARTS BIOT NUMBERS AND CHARACTERISTIC TIMES.
Outer Cu Solder
Inner Cu
Teflon
Biot #
3.3x10-6
5.0x10-5
3.3x10-6
6.4x10-2
-4
-3
-4
tc (s)
1.4x10
1.9x10
1.4x10
2.7x101
It is important to understand how these numbers will
affect the joint’s thermal performance. Materials with low
Biot numbers and short characteristic times will adjust
quickly to temperature changes. This will help shorten
heating and cooling cycles, however a material with a higher
characteristic time can store heat, thereby keeping the joint
close to melting temperature between cycles.
The amount of energy needed to melt the solder is
important given limited power. During the heating step, the
thermal energy that is put into the joint is distributed
between (i) raising the temperature of the joint components,
(ii) heat loss during melting (typically a fraction of the
melting energy and may assumed to be small), and (iii) the
energy, i.e. latent heat of fusion, that is required to induce
the phase change in the solder. Equation (7) may be used to
calculate the energy required to melt the solder in the joint.
The total heat required, Qt, is estimated from first principles
using the specific heat capacity for the materials, c, the mass
of the materials, m, and the change in temperature, ΔT. The
energy to achieve the phase change is calculated using the
material’s latent heat of fusion, Lf, and the mass of the
material.
(7)
Qt  (mc T  mL f ) solder  (mc T ) jo int materials
The temperature rise in each material depends on its
characteristic time scale. Those with time scales that are
orders of magnitude shorter than the cycle time will be
heated to at least the melting temperature of the solder
during the heating time. The temperature reached by
components with longer characteristic times depends upon
heat flow within the joint and is therefore difficult to obtain
without complex FEA/experimentation. We can, however,
rapidly obtain a lower bound upon the required energy by
calculating the energy required to melt the solder, Qsolder.
This may be estimated via (8). The results provide insight
into what is required to reduce the lower bound. For
example a reduction in the solder mass leads to a
proportional reduction in energy. We also desire to select a
solder that exhibits a low melting temperature, as well as
low specific heat and low latent heat of fusion.
(8)
Qsolder  (mcT  mL f ) solder
1684
For Squishbot1 we chose 60/40 solder as our starting
point because it is widely available and its properties are
well-known. This solder has a composition that is close to
the eutectic composition of lead-tin solders (63% tin and
37% solder). This yields one of the lowest melting points
for pure lead-tin solders, 190°C. Using (8), and the thermal
properties of 60/40 solder, we estimate that melting 0.5
grams of this solder requires 34 Joules. Given the heater
power, P, and the total melt energy, Qt, it is possible to
estimate melting time, t, via the relation in (9).
(9)
P  t Qsolder
Power limitations made it necessary to limit melting
power to 0.7W, which yields an unlocking time of at least
40 seconds. This made clear the need to alloy the solder and
thereby tune melting temperature for better performance.
The alloying process and other factors that influenced final
solder selection/design will be discussed in Section IV.
The time calculation in (9) assumes that 100% of the heat
is used to melt the solder. In practice, heat is lost to the
adjacent joint materials, therefore this estimate is a lower
bound on unlock time. A more detailed model would
consider the energy required to raise the temperature of each
component. Transient temperature simulations must be used
to determine temperature rise in components with
characteristic time scales that are approximately the same as,
or larger than, the length of the heat cycle.
The second part of the thermal analysis – cooling time –is
more complex given that the heat transfer out of the solder
depends upon the temperature of other components during
the cooling cycle. We may begin to understand where the
heat will travel by comparing the thermal resistances. As
there is a low convection coefficient for Squishbot1 the heat
transfer rate to the Teflon will be larger than the convection
transfer rate. This indicates that the temperature of the
Teflon will continue to rise a small amount during the
cooling step. To obtain a lower bound of the cooling time
we consider (i) the latent heat of fusion of the solder and (ii)
the energy that must be lost to decrease the temperature of
the solder below melting temperature. The heat of fusion
will dominate for small superheating. We may calculate a
cooling time for the joint for a given temperature of the
Teflon, as shown in (4).
In Squishbot1 the modeling results for lower bounds on
melting time indicated the need to use a solder alloy. The
thermal circuit modeling results highlighted the need for (i)
a high resistance material between the heater and Teflon to
reduce heat transfer to that material during heating and (ii) a
resistance on the Teflon side of the heater that is lower than
the convection resistance to help decrease cooling time.
Finally calculating the Biot and characteristic time scales
gives an insight into the materials behavior during cooling
and heating. Normally we would want to use materials with
low characteristic times to reduce cycle time. Given the
limited convection area, we obtain a counterintuitive result,
it works in our favor that the Teflon does not reach the
melting temperature and this helps reduce cooling time. This
type of insight demonstrates the power of first order models,
though more accurate modeling is still useful for fine tuning
of joint performance.
IV.
FLUID DESIGN
A. Solder Composition
In many designs, we have observed that the mechanism’s
components break before the solder joints. This ‘over
performance’ is useful as strength may be ‘traded off’ to
obtain more favorable thermal properties. There are many
alloying elements that may be added to a common lead-tin
solder to tune material properties. Table V shows the trends
in several key properties of the solder when a few common
alloying elements are added [7-9]. Table V portrays general
trends that are appropriate as guidelines for 1st order design.
More details may be found in the table’s references [7-9].
TABLE V TRENDS IN SOLDER PROPERTIES WITH ADDITION OF ALLOYS+
Element
Strength
Indium
↓
Bismuth
↓
Antimony (<6%)
↑
Silver (<5%)
↑
Ease of
Soldering
↓
↓
↓
↑
TMelt
↓
↓
≈
↑
Creep
Strength
↓
↓
↑
↑
+
Symbols indicate whether adding the element to lead-tin solders
increases (↑), decreases (↓), or has little effect (≈) on the properties.
For Squishbot1, we focused on Indium and Bismuth as
they lower the melting point of the solder. The wettability of
pure Indium and Bismuth to Copper is poor; therefore we
combined them with 60/40 solder to overcome this
limitation. We considered the alloys listed in Table VI.
TABLE VI POSSIBLE ADDITIONS TO LOWER 60/40 MELTING POINT.
Alloy
Composition by %
Chip Quik
10-30 In, 10-30 Pb, 7-13 Sn
Bismuth Alloy 52.5 Bi, 32 Pb, 15.5 Sn
Indalloy 117
44.7 Bi, 22.6 Pb, 8.3 Sn, 5.3 Cd
TMelt
58°C
95°C
47°C
B. Unlocking Temperatures
When the materials in Table VI are combined with 60/40
solder, the solder’s melting temperature decreases. The
unlocking temperature of the three custom mixes was
determined experimentally. A shear joint was created by
nesting one copper rod within a copper cylinder and then
soldering the two together. The joint was heated until it
unlocked under its own weight. The average radial gap
between rods was 175 microns and the axial overlap was
1.27cm. Two k-type thermocouples were used to measure
the temperatures on the (i) outer surface of the joint and (ii)
at the heater. A Mitutoyo 543 indicator was used to detect
joint motions that signaled an unlocking event.
Fig. 4 contains the temperatures of the heater and the
outer thermocouple when release of the inner joint was
observed. These results provide rapid, useful estimates of
melting temperature for first order performance models.
1685
Unlocking Temperature (⁰C)
Unlocking Temperature (°C)
heaters were capable of delivering 0.7 Watts of power into
each joint.
Unlocking Temperature of Shear Joints bonded with Different Solder Compositions
250
VI. COMMENTS ON THE NEED FOR SENSING
200
60/40 Solder
150
w/ Bismuth Alloy 203
100
w/ ChipQuik w/ Indalloy 117
50
0
0
1
Heater Temp.
Temperature sensors can reduce the cycle time by
ensuring that the solder is heated to just above the melting
point, thereby reducing the time/energy required to cycle the
joint. Currently Squishbot1 is equipped with thermistors;
however we are working to integrate polymer-based
piezoresistors and MEMS piezoresistors that gather thermal
(know if the PCM is solid or liquid) and strain information
(robot motion and displacement) within the robot.
2
3
4
Outer Joint Temp.
Fig.4. Measured unlocking temperatures with different alloying elements
C. Importance of Cycling
A thermally activated joint must be able to withstand
multiple cycles before de-wetting, delamination or crack
formation. These types of failures are not deterministic and
therefore investigated via experimental thermal cycling. For
example, Indium base solders may fail because of phase
segregation when there is an extreme and unidirectional
thermal gradient across the joint. Bismuth may become
brittle if it is solidified rapidly [7]. From experience, it is
known that the best means of ascertaining the risk associated
with these failures is to conduct failure tests. As part of the
Squishbot1 development, the joints were cycled 20+ times
to ensure suitable lifetime. During this testing, practical
issues with respect to other failure modes were discovered.
For example we found wiping of the solder off the locking
surfaces when the joint slid into/out of contact. We found
that smaller gaps and blunting modifications to the ‘wiping’
edges of the joint removed this failure mode. It is highly
recommended that designers plan to conduct cycling tests
early on in the prototyping phase to discover changes that
preserve expected performance or yield improvements in
performance.
D. Final Solder Selection
We elected to alloy the base solder with “Chip Quik” as
this lowered the melting temperature to a point that was
easily achieved with available heaters. Chip Quick is also
not toxic, unlike “Indalloy 117” which contains Cadmium.
V. RESISTIVE HEATERS
Resistive heaters are widely available in varied
geometries and capabilities. They are heated via Joule
heating that is induced by running a current through an
internal wire. The link between heat generation and current
is therefore easily known and the heater’s temperature is
easily set via current control. Cartridge heaters are too large
to use in our application, however small strain gages are
well-suited for use as miniature heaters. We used Vishay
062AK EA series strain gages with 120 Ohm resistance.
They have a footprint of 1.6x1.6 mm2. At each joint, we
placed two in parallel to reduce the total resistance to 60
Ohms and to enable us to heat the joint on both sides. These
VII. SQUISHBOT’S PERFORMANCE CHARACTERISTICS
Squishbot1 was able to locomote along its axis, steer left
and right, and lift its front while using one actuator and 3
solder-activated joints. The robot was able to crawl at 17.5
mm/s and steer at ~37 degrees/step as shown in Fig. 5. Post
optimization, joints require 15s cycle time to reconfigure.

A
B
C
Fig.5. Squishbot1 executing a turn. In A, the joints are configured in an
axial crawling state. The joint lock/unlock states are reconfigured to a
turning state between A and B, and then the robot turns between B and C.
REFERENCES
[1]
Shih-Chi Chen, “A Six-Degree-of -Freedom Compliant MicroManipulator for Silicon Optical Bench”, M.S. Thesis, MIT, 2003.
[2] Mark R. Jolly, Jonathan W. Bender, and J. David Carlson, “Properties
and Applications of Commercial Magnetorheological Fluids.” Journal
of Intelligent Material Systems and Structures, vol. 10, no. 1, pp. 5-13,
1999.
[3] Maria J. Telleria, Robert Panas, Martin L. Culpepper, “MR Fluid
Activated Spherical Joint for Precision Orientation Control,” 24th
American Society for Precision Engineering Annual Meeting, to be
published.
[4] Weijia Wen, XianXiang Huang, Shihe Yang, Kunquan Lu, and Ping
Sheng, “The Giant Electrorheological Effect in Suspensions of
Nanoparticles.” Nature Materials, vol. 2, 727-730, 2003.
[5] Aimee M. Ketner, Rakesh Kumar, Tanner S. Davies, Patrick W. Elder,
and Srinivasa R. Raghavan, “A Simple Class of Photorheological
Fluids: Surfactant Solutions with Viscosity Tunable by Light.”
Journal of the American Chemical Society, vol. 129, no. 6, 1553-1559,
2007.
[6] C. Ted Lee, Jr., Kenneth A. Smith, and T. Alan Hatton,
“Photoreversible Viscosity Changes and Gelation in Mixtures of
Hydrophobically Modified Polyelectrolytes and Photosensitive
Surfactants.” Macromolecules, vol. 37, no. 14, 5397-5405, 2004.
[7] Giles Humpston and David M. Jacobson, Principles of Soldering.
USA: ASM International, 2004.
[8] Howard H. Manko, Solders and Soldering, 2nd edition. USA:
McGraw-Hill Book Company, 1979.
[9] AWS Committee on Brazing and Soldering, Soldering Manual.
Miami, FL: American Welding Society, 1978.
[10] Mills, A. F., Heat and Mass Transfer, Concord, MA: Irwin Inc., 1995.
1686